Looking at the spectrogram of Hrec since April 1st, we noticed an increase of a sideband noise around the 50 Hz during May 2nd-4th, Figure1 (top). It seems to be correlated to a decrease of accuracy of DIFFp_TX (bottom).
As in #64018, in Figure 2 we observe two couple of sidebands probably due to a beating between the 50 Hz and DIFFp_TX at ±0.75 Hz and ±0.42 Hz (roughly).
A residual of these sidebands seems often present: purple curve of Figure 2 and spectrogram in Figure 1.
We performed some measurements with a portable magnetometer: 3-axial magnetic probe FL3 100 Stefan Mayer.
Figure 1 compares spectra recorded in two locations:
The GUARD3 spectra clearly show the noise: 0.23 Hz spaced lines peaking at 8Hz. The peak amplitude (spectra are obtained with 25s window) is approx 2 nT/sqrtHz. Which is a factor approx 100 larger than measured in the location EXTMAG, but similar to that measured by the magnetometers on the NI and WI tower base. Might be due to telluric currents in the chambers?
As a check of the probe calibration, in Figures 2 and 3 we compare the EXTMAG measurement with that of the buried magnetometer, ENV_EXT_MAG_N. The fluxgate probe is more noisy (intrinsic noise) but around 8Hz they are measuring the same.
The indication is that the magnetic disturbance is radiated from the methan gas pipelines. A noise clearly associated to the power supply for galvanic currents in these pipelines was observed in the year 2022 (55542 and 57973). At that time the noise was shaped as approx. 5Hz comb.
Federico has contacted the methan gas company (SNAM) to investigate further the source.
Today at 13:00 LT we could see an airplane flying over the main building, probably just departed from the lane nearby; one minute later it can be spotted in the CEB microphone signals, but it did not have an impact on Hrec; some 25 minutes later it probably came back along the same route, luckily causing no effect again on the sensitivity.
Glitches disappeared around 6:45 UTC. Has some action been taken? Indeed, square waves were seen also in the CEB UPS CURR T monitor, as for the old problem with WI heating belts that Bas pointed out.
A new noise is seen in the external magnetometer starting April 22 19:56 UTC. Figures 1 and 2. The noise has been detected and studied by Renato and Federico. Here is a report of the investigation so far.
The noise is impulsive, as seen in Figures 3,4,5. Observed in the time domain it consists of a carrier at about 8Hz switching on-off with a cycle of approximately 4.4s (0.23 Hz): on for 2.2s and off for 2.2s.
The noise is seen by all magnetometers site-wide. Internal magnetometers sees it a factor up to about 10 larger than the external "N" magnetometer, in Figure 6 and Figure 7 (the EXT_N magnetometer is the gray spectrum in all 3 plots). Magnetometers that see it a bit more intense are those close to NI, WI and WE towers.
A close look to the narrow spectral structure of the noise shows that it does not follow the mains, in Figure 8.
These hints suggest a DC-powered PWM source (Federico).
Bruco runs using ENV_NI_MAG_V as target, and EXCLUDING all magnetometers (ENV_*MAG*) channels, show that this disturbance is not coherent with any other ENV channel (https://scientists.virgo-gw.eu/DataAnalysis/DetCharDev/bruco/users/fiori/ENV_NI_MAG_V_1397853318_ENV/) with the exception of a small coherence with the ENV_CEB_ELECTRIC. In particular no coherence, nor evidence of the noise is seen in all IPS and UPS probes. No coherence is also measured with any other Virgo signal (https://scientists.virgo-gw.eu/DataAnalysis/DetCharDev/bruco/users/fiori/ENV_NI_MAG_V_1397853318/). While s is a good news for the interferometer, it does not help much locating the source.
Measurements will be done also to check for possible external sources, like it was for galvanic currents in the natural gas pipelines (https://logbook.virgo-gw.eu/virgo/?r=55542).
We performed some measurements with a portable magnetometer: 3-axial magnetic probe FL3 100 Stefan Mayer.
Figure 1 compares spectra recorded in two locations:
The GUARD3 spectra clearly show the noise: 0.23 Hz spaced lines peaking at 8Hz. The peak amplitude (spectra are obtained with 25s window) is approx 2 nT/sqrtHz. Which is a factor approx 100 larger than measured in the location EXTMAG, but similar to that measured by the magnetometers on the NI and WI tower base. Might be due to telluric currents in the chambers?
As a check of the probe calibration, in Figures 2 and 3 we compare the EXTMAG measurement with that of the buried magnetometer, ENV_EXT_MAG_N. The fluxgate probe is more noisy (intrinsic noise) but around 8Hz they are measuring the same.
The indication is that the magnetic disturbance is radiated from the methan gas pipelines. A noise clearly associated to the power supply for galvanic currents in these pipelines was observed in the year 2022 (55542 and 57973). At that time the noise was shaped as approx. 5Hz comb.
Federico has contacted the methan gas company (SNAM) to investigate further the source.
As observed by Federico, the metronix magnetometer in MCB is having an issue starting from 2023 October 2 UTC 00:03:03 (Figure 1). The signal looks corrupted (Figure 2). This does not seem an issue with the ADC and neither with the driving electronics, since other channels sharing the same boards are fine. We will investigate further at next maintenance.
This fauty condition was not signalled by the DMS DeadChannel since the associated rms was smaller than the set threshold. The rms threshold has been updated (0.35 to 1 nT) and this channel temporary shelved.
This is a first (preliminary) look at the effect of acoustic injection into Hrec, performed on Thursday 18.
For each injection (at the 0.03, 0.04 and 0.05 V levels) the BNS range decreased.
Here the entire Hrec range from 10 to 1000 Hz.
There are regions growing at low frequency (around 120 and 170 Hz) and high frequency (the most relevant from 500 to 540 Hz).
Below 100 Hz nothing seems to have changed significantly.
We tested the sensitivity of the top-stage BS accelerometers at three different levels of acoustic injection (nominally 0.03 - 0.04 - 0.05 V).
Noise was filtered from 8 to 2000 Hz, but the actual noise produced was only from 20 Hz and up, as the speaker is unable to go lower.
The fourth injection (18:04:40 UTC - (80 -100) Hz, 300 sec, 0.04 V) aimed to produce the same level around the frequency of ~93 Hz (without generating out-of-band noise) where one of the horizontal BS F0 accelerometers (Sa_BS_F0_ACC_H3_500Hz) is usually excited during acoustic injections, in order to check whether the coupling is linear or not.
Unfortunately, the filter was too narrow, and the noise spectral density at 93 Hz was lower than expected.
However, we can say that generating a higher level by a factor of 2 (with no noise at lower frequencies) resulted in an increase of the peak by about a factor of 2 (difficult to estimate accurately). It appears that there is no upconversion and the system is "linear."
At first glance one might think of general coupling due to the external accelerometer cables running not too far from the magnetometers (as indeed they are), but there are high-level signals in the acceleration spectrum that have no coherence. If it were a coupling from the cable, it should be broadband.
We performed a set of four acoustic injections in CEB with the AcousticInjectionCEB VPM process:
1) 17:07:55 UTC - (8 - 2000) Hz, 600 sec, 0.03 V
2) 17:21:05 UTC - (8 - 2000) Hz, 600 sec, 0.04 V (25 min glitch at ~17:29:24 UTC)
3) 17:33:48 UTC - (8 - 2000 )Hz, 600 sec, 0.05 V
4) 18:04:40 UTC - (80 -100) Hz, 300 sec, 0.04 V
The directory path of output files is /virgoData/NoiseInjections/AcousticInjectionsO4/output. The list of file injections is reported below:
1) AcousticColored_NOISE_CEB_DetTerrace-1397495261.txt
2) AcousticColored_NOISE_CEB_DetTerrace-1397496052.txt
3) AcousticColored_NOISE_CEB_DetTerrace-1397496814.txt
4) AcousticColored_NOISE_CEB_DetTerrace-1397498666.txt
As Piernicola confirmed, the sidebands at 46.7 Hz and 53.3 Hz are the beating between 50 Hz and DIFFp_TX line at 3.3 Hz (purple curve).
Indeed, when the DIFFp_TX line was set to 4.8 Hz, there were two sidebands at 45.2 Hz and 54.8 Hz, respectively (blue curve).
Looking at the spectrogram of Hrec, we observed a peak at 47.5 Hz which arose on April 15th, ~01:00 UTC (between two locks), see Figure 1 and 2.
It could be related to the F7 crossbar mode of the NI tower that we measured by magnetic injections to be at 47.65 Hz (VIR-0300A-24).
In this case, it seems that the line has shifted slightly in frequency.
Continuing the investigation of the spectral noise on the first week of O4b data, we found an interesting structure of "wandering sidebands" around the 50 Hz line. (S
Figure 1 shows the spectrogram in the region 42-58 Hz around the frequency of the electric mains. Two pairs of wandering sidebands are visible at 44 (Figure 2) and 56 Hz, and at 46.7 (Figure 3) and 53.3 Hz (Figure 5).
The oscillations in frequency, especially those of the 44 and 56 Hz lines, are the same as those of the 50 Hz line (Figure 4) but enlarged by many (~20) times: ~0.50 Hz vs ~0.025 Hz
Additionally, taking advantage of a couple of glitches present in the analyzed data period (the vertical lines), we notice that the 44 and 47.3 Hz lines are synchronous with the 50 Hz line. Also, the 53.3 Hz line seems to have the same phase, while the one at 56 Hz is flipped vertically: Figure 5.
The search for bilinear couplings with MONET highlighted that the 46.7 and 53.3 Hz sidebands are produced by the modulation of the mains line (ENV_CEB_UPS_VOLT_R_2000Hz) and the ASC_DIFFp_T{X,Y} angular control channels: Figure 6 and text file.
A barely negligible coherence has been found for the lines at 44 and 56 Hz; just a 0.1 with SDB1_LC_TX at 56 Hz: Figure 7. This could be a consequence of the large variations in the frequency of the line, which dilute the cohere over multiple frequency bins and, averaged gets small.
Of course, the same structure is present around the other harmonics of the 50 Hz line.
As Piernicola confirmed, the sidebands at 46.7 Hz and 53.3 Hz are the beating between 50 Hz and DIFFp_TX line at 3.3 Hz (purple curve).
Indeed, when the DIFFp_TX line was set to 4.8 Hz, there were two sidebands at 45.2 Hz and 54.8 Hz, respectively (blue curve).
ERRATA CORRIGE: the sidebands at 44 and 56 Hz are non-physical. These are just an aliasing artefact, consequence of an improper chunk-by-chunk downsampling to produce the resulting spectrogram.
Figure 1: fixed version of the spectrogram in the range 42-58 Hz.
Figure 2: animated GIF that compares the correct spectrogram and the wrong one in the main entry.
The 50 Hz sidebands at 0.75 Hz are modelled by the bilinear coupling of ENV_CEB_UPS_VOLT_R_2000Hz and ASC_DIFFp_TX, as visible in the daily MONET results of May 2. The modulator ASC_COMMp_TX produces interesting structures too, although just below the default coherence threshold set on the daily MONET analysis.
Looking at the spectrogram of Hrec since April 1st, we noticed an increase of a sideband noise around the 50 Hz during May 2nd-4th, Figure1 (top). It seems to be correlated to a decrease of accuracy of DIFFp_TX (bottom).
As in #64018, in Figure 2 we observe two couple of sidebands probably due to a beating between the 50 Hz and DIFFp_TX at ±0.75 Hz and ±0.42 Hz (roughly).
A residual of these sidebands seems often present: purple curve of Figure 2 and spectrogram in Figure 1.
On April 9 we performed a set of short (60s) colored acoustic injections in the CEB. The ITF was locked in LOWNOISE 3 since a few minutes.
The purpose was to tune the level of the injection in such a way to avoid triggering the 92 Hz oscillation of the BS F0 accelerometers (mostly Sa_BS_F0_ACC_H3) which caused unlocks in the past (https://logbook.virgo-gw.eu/virgo/?r=63709). The purpose of the present test was not to understand the nature and physical reason of the oscillation onset, which remains to be understood.
We injected colored noise between 20 and 2000Hz with 3 levels: 0.03 V, 0.04 V and 0.05 V, one 25min glitch occurred right at the beginning of the 0.05 V injection (Figure 1).
The intensity on the acoustic noise produced in the hall, measured in the position of the hall microphone, was of the order of a factor 20 above the quiet condition (Figure 2). The 92 Hz oscillation of the BS F0 ACC H3 accelerometer onsets already with the 0.03 V but amplifies x5 and x10 when the injected level increases from 0.03V to 0.04V and from 0.03V to 0.05V, respectively (Figure 3).
Figure 4 shows the impact on hrec. The noise in hrec increased at some structures, mostly known to be associated to SDB1. These excitations are well visible with all injected levels (Figure 5, which also shows the 25min glitch).
It seems to notice also an increase of the 74 Hz line (DARM loop gain) and the 227 Hz line (SSFS loop gain). As well, it seems to notice a small increase of the noise in the bucket. It seems that particularly the injection at 0.05 V, brings the ITF to a different working point where other noises (e.g. frequency noise) are more coupled.
However, spectrograms in Figures 6 and 7 indicate that the amplitude of 74 Hz and 227 Hz lines was not so stable during the injection period (we were at the beginning of a lock) and its correlation with single acoustic injections is not straightforward.
Paolo's suggestion is to better study this behaviour with longer injections.
The issue with the extra spike at the beginning of the colored noise injection was related to time settings in the ACL commands in the ENVnoise process. These modifications were made by Alain (on March 26):
Figure 1 shows the result of the test: blue is with the OLD settings, purple is the with the NEW settings (the same colored noise with amplitude 0.025 V has been injected in both cases). The spike is no longer present.
The same test was done by injecting the colored noise from the ENVnoise and also using the ENV_MAIN node which calls the AcousticInjectionCEB process.
The increased coherence with ENV_NEB_IPS_CURR seems associated to the accidental presence of a glitch in hrec (one of the 25-min glitches) and a glitch in the ENV_NEB_IPS_CURR sensors.
The two glitches are NOT coincident indeed but a few sec away (the IPS glitch occurs after), but they both fall in the 10s window of the Bruco computation.
Figure 1: coherences are computed 120s before the glitch (purple) and in a 120s window containing the glitches.
Figure 2: spectrograms show that the two glitches are a few seconds away. The COHETIME plot implemented in dataDisplay measures a sort of "anti-coherence" (the coherence goes to a lower value in the time window that contains both glitches).
It is not clear what is peculiar to the NEB_IPS_CURR sensor glitch to produce coherence with Hrec?
Today we have concluded the tuning of DMS threshold for a few ENV channels which have been observed to fluctuate between green and red inside the DeadChannel monitor.
In particual for channel ENV_IB_CT_FINGER_ACC_Y we had to first fix the issue with too large 50 Hz and harmonics. The same problem was obseved in the past 62620 and fixed by replacing the BNC to lemo adaptor. Figure 1 shows the spectrum of the accelerometer powered off (blue) with the old connector, and Figure 2 shows the same with the new connector. Figure 3 shows the procedure followed for setting threshold for the sensor ENV_IB_CT_FINGER_ACC_Y.
On Friday afternoon, while checking the power supply of some accelerometers on the DET terrace, we noticed that the flashing yellow light of the LargeCoil was ON. Indeed, the power amplifier was ON (but without any signal sent from the DAC).
Trying to turn it OFF via the Process Online Monitoring page, we found the LargeCoil_CEB gray (NEB and WEB were correctly green). The process could not be restarted. The box remained gray and the relay powering the amplifier remained in the "1" state.
Today I tried again from remote and the process restarted. I turned off all the relays at 10:51:10 UTC - There is a huge spike on the monitoring channels (ENV_CEB_MAG_COIL_CURR_MON and ENV_CEB_MAG_COIL_VOLT_MON), but it is a fake signal due to the fact that these signals are coming from the power amplifier that turned off (nothing is visible in the CEB magnetometers and in Hrec).
We will check the system again during commissioning (Monday afternoon), also because it seems that the status of the relays visible to the right of the green box is always zero, even though they are (really) set to 1.
On the March 12th, a shaker has been installed on the Laser Bench to perform sismic noise injection. After those studies done on the 14th, the electronics that controls the shaker (deployed on the ground) has been left ON (with 0V of input voltage) to have the possibility to further perform other remote injections.
However, entering in the laser lab few days later, we noticed a slightly higher acoustic noise than usual. We decided then to switch OFF that electronics, and indeed the noise reduced.
Looking at the environmental sensors present in laser lab, we found this noise visible as both high frequency broad noise (>200Hz) and with some lines a low frequency (~45Hz and harmonics).
Moreover, we see it on acoustic (see fig. 1-3, 5) and sismic (fig. 4) sensors. While the noise is visible on both benches, the effect (seismic and acoustic) on EIB is really faint (see fig 4-5).
Nevertheless, we didn't find any noticeble effect on Hrec.
Extra hrec noises were present during tonight locks (Figures 1 and 2):
We do not know it the two are related.
Investigating we found that the PSTAB0 line at 43 Hz is injected. and it is seen in various INJ signals (Figure 3).
The plan is to try to switch it off after locking and check the impack on hrec.
PSTAB0 line (43Hz) has been switched off (amp=0) at 12.52.40UTC.
We could then relock in LN3: the bump at 43 disappeared while the broadband noise was still there.
To be noted that PSTAB1 line (236Hz) is still injected while, as far as I know, not used by any loop.
I have been able to run the noise budget by hand, after changing some of the Pcal channels used that have disappeared in the last two days from the raw data.
Figure 1. Shows that the noise budget thinks the OMC projection is high. Normally it is 2 orders of magnitude below the sensitivity curve.
Figure 2 shows that the OMC error signal spectrum is 3 orders of magnitude higher than usual. The sensing noise is three orders of magnitude higher.
Figure 3. In the lock of ~17:00 UTC the problem is no longer present. Maybe the issue was removed by the restart of the SDB1_OMC process that was done today by Alain around 15:00 UTC.
The issue was due to the release v1r22p17 of Acl . A Cm command related to the ACL_SUM_CH object use to select the error signal was not applied .
Below the messages from the SDB1_OMC logfile
2024-03-25-23h01m05-UTC>WARNING-AcSumChCmSet> OMC1_in1 - missing arguments: cmType "AcSumChSet" -t -d -d -t ... - rtn -1
2024-03-25-23h01m05-UTC>WARNING-AcSumChCmSet> OMC1_in0d5 - missing arguments: cmType "AcSumChSet" -t -d -d -t ... - rtn -1
As a result , the error was not the expected one and as consequence a large correction signal.
The large broadband noise up to some hundred Hz is highly coherent between Hrec and V1:DET_B1_DC, as a result of a dedicated BruCo run: link (I'm currently rerunning the script to fix the cropped y-scale). In the table, you can find also a bunch of channels coherent with the large peak at 43 Hz.
Figure 1: coherence and noise projection of V1:DET_B1_DC to Hrec.
[For some reason, this message has remained in the Drafts, even if I received the notification that it was added to the Logbook...]
Ignore my previous entry. The highlighted channel is one of those recently added to DET, as reported in this logbook entry: #63738. I'm adding them to the excluded channels from BruCo analysis:
/data/dev/detchar/online/bruco/share/virgo_excluded_channels_Hrec_hoft.txt
EDIT: BruCo daily results for Hrec (link) have correctly excluded this channel from the analysis. Unfortunately, I forgot to edit the corresponding list for DARM, and the coherence results for the last day (link) have been dominated by the new DET channel: ignore this channel from the BruCo results.
Now, I have edited the excluded channel list for DARM too. I've opened this git issue to keep track of the various excluded channel lists and keep them up to date.
This is a follow up of https://logbook.virgo-gw.eu/virgo/?r=63649
Friday afternoon we used two hours of commissioning to test the ENV_MAIN metratron node. The ITF was locked in LOW_NOISE_3_SQZ, COMMISSIONING.
WE performed this sequence of environmental injections:
The magnetic injection was succesfully performed at CEB, NEB and NEB (Figure 1). The 9V amplitude although close to the limits for current (6Apk) and Voltage (70Vpk) did not saturate the amplifier (Figure 2). With this amplitude the signal is seen sufficiently well in hrec for NEB and WEB up to about 100Hz. For CEB the signal in hrec looks not so loud (Figure 3) but we might gain a bit by doubling the injection time to 600s. For NEB and WEB the Bartington magnetometers ENV_NE(WE)_MAG_V,N,W placed close to the NE(WE) tower have to be used for the coupling function computation, because the metronix magnetometers on NEB and WEB saturate for the low frequency part of the sweep.
The output files were successfully produced in the designated directory: /virgoData/NoiseInjections/MagneticInjectionsO4/output/ and /virgoData/NoiseInjections/AcousticInjectionsO4/output/
One thing that is useful to be implemented in is that once the sweep injection is concluded a few seconds are waited before switching off the amplifier.
Instead the acoustic injection did not preform as expected because the large spike noise at the beginning of the injection appeared again (!) and of course unlocked the ITF: Figure 4
Indeed, to our understanding this should not have happened since the used
/virgoDev/Automation/userapps/ENV_MAIN.py
now points to the patched version of the code (v2r1p1), which when tested by launching the AcousticInjectionCEB process from the VPM performed ok (i.e. no spike)
.... reading from the ENV_MAIN.py:
# noise_injection_script = "/virgoDev/EnvSwepts/v2r2/Linux-x86_64-CL7/bin/AcousticColoredInjection-conda /virgoData/NoiseInjections/ENV/AcousticInjectionCEB.cfg"
noise_injection_script = "/virgoApp/EnvSwepts/v2r1p1/Linux-x86_64-CL7/bin/AcousticColoredInjection-conda /virgoData/NoiseInjections/ENV/AcousticInjectionCEB.cfg
We need Franco or Rhys to have a look into this.
The issue with the extra spike at the beginning of the colored noise injection was related to time settings in the ACL commands in the ENVnoise process. These modifications were made by Alain (on March 26):
Figure 1 shows the result of the test: blue is with the OLD settings, purple is the with the NEW settings (the same colored noise with amplitude 0.025 V has been injected in both cases). The spike is no longer present.
The same test was done by injecting the colored noise from the ENVnoise and also using the ENV_MAIN node which calls the AcousticInjectionCEB process.
On April 9 we performed a set of short (60s) colored acoustic injections in the CEB. The ITF was locked in LOWNOISE 3 since a few minutes.
The purpose was to tune the level of the injection in such a way to avoid triggering the 92 Hz oscillation of the BS F0 accelerometers (mostly Sa_BS_F0_ACC_H3) which caused unlocks in the past (https://logbook.virgo-gw.eu/virgo/?r=63709). The purpose of the present test was not to understand the nature and physical reason of the oscillation onset, which remains to be understood.
We injected colored noise between 20 and 2000Hz with 3 levels: 0.03 V, 0.04 V and 0.05 V, one 25min glitch occurred right at the beginning of the 0.05 V injection (Figure 1).
The intensity on the acoustic noise produced in the hall, measured in the position of the hall microphone, was of the order of a factor 20 above the quiet condition (Figure 2). The 92 Hz oscillation of the BS F0 ACC H3 accelerometer onsets already with the 0.03 V but amplifies x5 and x10 when the injected level increases from 0.03V to 0.04V and from 0.03V to 0.05V, respectively (Figure 3).
Figure 4 shows the impact on hrec. The noise in hrec increased at some structures, mostly known to be associated to SDB1. These excitations are well visible with all injected levels (Figure 5, which also shows the 25min glitch).
It seems to notice also an increase of the 74 Hz line (DARM loop gain) and the 227 Hz line (SSFS loop gain). As well, it seems to notice a small increase of the noise in the bucket. It seems that particularly the injection at 0.05 V, brings the ITF to a different working point where other noises (e.g. frequency noise) are more coupled.
However, spectrograms in Figures 6 and 7 indicate that the amplitude of 74 Hz and 227 Hz lines was not so stable during the injection period (we were at the beginning of a lock) and its correlation with single acoustic injections is not straightforward.
Paolo's suggestion is to better study this behaviour with longer injections.
We performed a set of four acoustic injections in CEB with the AcousticInjectionCEB VPM process:
1) 17:07:55 UTC - (8 - 2000) Hz, 600 sec, 0.03 V
2) 17:21:05 UTC - (8 - 2000) Hz, 600 sec, 0.04 V (25 min glitch at ~17:29:24 UTC)
3) 17:33:48 UTC - (8 - 2000 )Hz, 600 sec, 0.05 V
4) 18:04:40 UTC - (80 -100) Hz, 300 sec, 0.04 V
The directory path of output files is /virgoData/NoiseInjections/AcousticInjectionsO4/output. The list of file injections is reported below:
1) AcousticColored_NOISE_CEB_DetTerrace-1397495261.txt
2) AcousticColored_NOISE_CEB_DetTerrace-1397496052.txt
3) AcousticColored_NOISE_CEB_DetTerrace-1397496814.txt
4) AcousticColored_NOISE_CEB_DetTerrace-1397498666.txt
We tested the sensitivity of the top-stage BS accelerometers at three different levels of acoustic injection (nominally 0.03 - 0.04 - 0.05 V).
Noise was filtered from 8 to 2000 Hz, but the actual noise produced was only from 20 Hz and up, as the speaker is unable to go lower.
The fourth injection (18:04:40 UTC - (80 -100) Hz, 300 sec, 0.04 V) aimed to produce the same level around the frequency of ~93 Hz (without generating out-of-band noise) where one of the horizontal BS F0 accelerometers (Sa_BS_F0_ACC_H3_500Hz) is usually excited during acoustic injections, in order to check whether the coupling is linear or not.
Unfortunately, the filter was too narrow, and the noise spectral density at 93 Hz was lower than expected.
However, we can say that generating a higher level by a factor of 2 (with no noise at lower frequencies) resulted in an increase of the peak by about a factor of 2 (difficult to estimate accurately). It appears that there is no upconversion and the system is "linear."
At first glance one might think of general coupling due to the external accelerometer cables running not too far from the magnetometers (as indeed they are), but there are high-level signals in the acceleration spectrum that have no coherence. If it were a coupling from the cable, it should be broadband.
This is a first (preliminary) look at the effect of acoustic injection into Hrec, performed on Thursday 18.
For each injection (at the 0.03, 0.04 and 0.05 V levels) the BNS range decreased.
Here the entire Hrec range from 10 to 1000 Hz.
There are regions growing at low frequency (around 120 and 170 Hz) and high frequency (the most relevant from 500 to 540 Hz).
Below 100 Hz nothing seems to have changed significantly.
Yesterday morning we noticed a strange behavior of the Etalon loop, both on NI (fig. 1) and WI (fig. 2) tower, probably due to the restart of some ACL servers that generate the signal for the control loop while the loop was still ON.
In particular, the loop seemed to have an inverted sign as it was heating the tower when the RH probe temperature was above the setting point. A stop/start of both loops (NI: 15h53m-UTC; WI: 17h57m-UTC) seemed to solve the problem (see fig. 3). However, as observed by Michal, it is possible that this huge and fast change of temperature of the two towers (fig. 4) had an effect on the lock stability and on the sensitivity.
Due to the slow control of the loop, the effect of the issue were visible with a huge delay; we then decided to postpone the start of the etalon scan, expected for the beginning of the ER. From the data of this morning, I would suggest not to start the scan before the end of the day in order to start with a stable initial state.
Beside, it is worth to notice also a strange behavior of the NI RH temperature probe. In figure 5 we can observe that after the lost of data due to the server restart of tuesday morning (between 7:10 and 8:30 UTC), the temperature read on the NI RH did a jump (and the same did the NI etalon err signal), while on the WI its behavior was as expected.
Figure 1. The NI etalon has nicely stabilized over the past few days to fluctuations of ~20mK peak to peak, but the WI etalon fluctuations are not decreasing, with fluctuations remaining at the ~100mK peak to peak. Is there a need to change the WI etalon control filter parameters so it moves also from on/off regime to a linear control regime?
We concluded the campaign of near field magnetic injections of elog 63540:
We performed the full set of environmental injections by using the ENV_MAIN node (ITF was in LOW NOISE 3 SQZ)
We tried to perform the "same" colored noise injections in CEB by using the "AcousticInjectionCEB" process in the VPM (pointing to the new executable /virgoApp/EnvSweepts/v2r1p1/... )
Conclusion: we need to test again the ENV_MAIN after these fixings. About 1 hour is needed.
"NE mysterious noise"
One additional event occurred March 16 at 22:23 UTC and caused an unlock: Figure 1 (trend), Figure 2 and Figure 3 (spectrogram of accelerometers). The frequency of the turbo pump line (547Hz) did not seem to be affected at all (Figure 4).
At that time, according to Antonio's and Fabio's reports: the main turbo pump was switched off and the spare turbo pump was on, the accelerometer named ENV_NE_BOTTOM_ACC was placed on the main turbo (see picture in Figure 5).
Figure 6 compares the spectrum of this accelerometer (orange) with similar accelerometers on the NE big flange (blue) and on the NE cryo trap (green). The signal seen by the accelerometer on the main pump indeed detected a large signal.
We moved the accelerometer named ENV_NE_MIDDLE_ACC on top of the spare turbo pump. See picture in Figure 7.
We switched off the external cooling fan of the main turbo.
Extra glitches in accelerometer ENV_NE_ACC_Z: in Figure 2, we notice several fast glitches which are not seen by the other accelerometer. Need further investigation.
Following the investigation in the entry #63491, this morning we perfomed a set of noise injections on both laser bench and external injection bench.
We started with the injection on laser bench by means of shaker installed few days ago #63577.
LB NOISE INJECTIONS
ITF locked
DAC channel: ENV_NOISE_CEB_EEroom
Freq. line 17.9 Hz, A=0.1 V, 09:59:55-10:13:57 UTCFreq. line 8 Hz, A=0.7 V, 10:37:20-10:42:46 UTCFreq. line 8 Hz, A=0.6 V, 10:50:58-10:57:20 UTCSecond part of the shift ITF unlocked (INJ standalone)Quiet = 15:35:00 (5’) Freq. line 17.9 Hz, A=0.1 V, 15:40:20-15:48:00 UTCFreq. line 8 Hz, A=0.7 V, 15:49:51-15:58:36 UTCFreq. line 8 Hz, A=0.6 V, 16:00:04 - 16:06:01 UTCColored noise (1-15) Hz: 16:18:27 - 16:23:39 UTC
EIB NOISE INJECTIONS
ITF unlocked (INJ standalone)
(previous EIB injections #61713)
We injected a sine of 0.5Hz frequency (Henk Jan adjusted the cfg file to have the right amplitude by just enabling sine noise button)
| gps start | DOF | amplitude | duration |
| 16:24:26 | z | 10 um | 3 min |
| 16:28:08 | y | - | - |
| 16:33:31 | x | - | - |
| 16:37:46 | tx | 10 urad | - |
| 16:41:26 | ty | - | - |
| 16:44:45 | tz | - | - |
The analysis of these injections will follow.
During the maintenance, we installed a shaker on the laser bench, see the attached Figures 1 and 2.
We checked the optimal parameter configuration to have the maximum displacement of the bench:
- freq. line injected at 17.9 Hz (laser bench horizontal resonance)
- voltage amplitude 0.1 V.
Looking at the episensor ENV_LB_SEIS_*, we achieved a spectral displacement of the laser bench at 17.9 Hz:
| channel direction | ||
| x | ~1e-2 | ~0.8 |
| y | ~1e-3 | ~0.08 |
| z | ~4e-2 | ~3.2 |
We checked that this injected line was not visible on EIB, Figure 3.
We also tried to inject lines in the freq. range (1-5) Hz but the maximum displacement achieved was lower by a factor of ten.
On the March 12th, a shaker has been installed on the Laser Bench to perform sismic noise injection. After those studies done on the 14th, the electronics that controls the shaker (deployed on the ground) has been left ON (with 0V of input voltage) to have the possibility to further perform other remote injections.
However, entering in the laser lab few days later, we noticed a slightly higher acoustic noise than usual. We decided then to switch OFF that electronics, and indeed the noise reduced.
Looking at the environmental sensors present in laser lab, we found this noise visible as both high frequency broad noise (>200Hz) and with some lines a low frequency (~45Hz and harmonics).
Moreover, we see it on acoustic (see fig. 1-3, 5) and sismic (fig. 4) sensors. While the noise is visible on both benches, the effect (seismic and acoustic) on EIB is really faint (see fig 4-5).
Nevertheless, we didn't find any noticeble effect on Hrec.
